Jim Dettman
jimdettman at verizon.net
Thu Sep 11 16:12:25 CDT 2014
Don, Yes it makes sense, although it can lead to a very complex design. For example, there are different types of samples and every test defined may not apply. If we had an air sample, ppm of a contaminate would, but a slump test would not (it's a test for concrete). So you'd need to add a table for sample types, and then an allowed test table for each type. But that is indeed the correct way to do it. If you've modeled a process properly, then I should be able to use the system without making changes to the database itself, only the data it contains. A new test coming along should not mean that I need to add a new table, queries, reports, etc, but just more records. But this brings up one other point with these types of systems; do you fully build in the knowledge of what is related to what or no? An example is a configurable assembly where parts A and B might go together, but C and D cannot. When you start building in those types of rules, then things get very complex. In general I find client never want to go that far, and they leave it up the user to determine if a "slump test" can be performed against an air sample. As far as the rest: <<Second, and more interestingly, where does one draw the line between delivering exactly (or as near to it) what the client has asked for, and designing for a future that even the client may not anticipate?>> That's a large gray area, but one of the things a client is paying you for is your advice on how it should be done. I always like to sit down and explain what the ramifications are of doing it one way vs another, especially if they want to take short cuts. In the end, it's their choice or course. <<Is it ever appropriate to "out-think" the client and deliver a system that they didn't ask for and don't know they'll be needing?>> I think so. Again, their paying you to know what needs to be done. I wrote a small system for a health care facility about six years ago. They thought I was going way over board with all the stuff I brought up, but there have been no changes since it went in. In fact they called just a few weeks ago about adding a new nurse and that was the first time I've heard from them in a couple of years (they forgot where it was in the menus). It's tracks patients, services, vaccines, billing etc and has a series of reports to monitor all that. I asked how it was working and if anything was giving them a problem. She said "no, it's done everything we've needed it to do". That's when you know you've done a design right. Jim. -----Original Message----- From: accessd-bounces at databaseadvisors.com [mailto:accessd-bounces at databaseadvisors.com] On Behalf Of McGillivray, Don Sent: Thursday, September 11, 2014 01:27 PM To: Access Developers discussion and problem solving Subject: [AccessD] Design Considerations - Was: Table Structure Ideas The direction of this thread interests me because it resembles my own thought processes when evaluating different design considerations. I don't want to derail the conversation or muddy the water for the original poster, but I'd be interested in hearing the opinions of others on the implications of proposing a design approach that exceeds a client's stated requirements. In my experience, the client often has only a vague idea about how to structure data related to some task or problem, and usually has some narrow view about what a user interface should look like and how it should behave. Left to their own devices, clients would structure the data AFTER designing their preferred UI. They'll force the data structure to fit the UI, instead of ensuring a sound structure. Then, because the system was designed with a narrow vision of its function, entirely new systems may have to be developed (or existing ones extensively modified) to accommodate modest (and foreseeable) changes in the operating environment. (e.g., now we're testing samples of material x, using tests y and z.) This often leads to systems performing similar work in separate silos with no integration between them. How many times have we seen this when called in to design a solution? In the case of the problem presented here, it seems to me that there are three main entities that need to be modeled: Samples, Tests, and Test Results. In reality, there may be other entities involved with the application, such as the Client, the Job, the Sample Site, etc., but for now let's stick with Samples, Tests, and Results. Given this problem, I'd be inclined to construct a table for each of those entities, such that each sample may be related to one or more tests, and each test may be related to one or more results. A subsidiary table (or tables) could contain definitions for the various tests and constraining the types of results relevant to each test, and the types of tests relevant to each sample type. The advantage to this approach is that it provides a foundation for a solution to the current problem, while accommodating the (almost) inevitable future need to deal with new types of samples, tests, and results. Rather than adding new tables to accommodate ne! w tests and their results, we just add new records to the subsidiary tables that define the attributes and relationships of each. Test A may be materially different from Test B, in terms of how it's conducted and what is measured, but at bottom they're identical: activities that yield results that need to be recorded and evaluated. Now, I know that this sort of a design is a bit more complex, in terms of its development, implementation, and ongoing maintenance. It's not as easy to explain to the client, either. So I'd also be tempted to round off some corners and compromise in the directions so far proposed. So, I have a couple of questions: First, does the approach described above make sense, or am I totally off-base? I'm always impressed with the quality of the contributions from AccessD posters, and, not having seen a suggestion resembling mine, I wonder if I'm over-complicating things or missing something obvious that would argue against it. I often think in this way about a problem when designing a solution, and if there's a better way, I'm all ears. Second, and more interestingly, where does one draw the line between delivering exactly (or as near to it) what the client has asked for, and designing for a future that even the client may not anticipate? Is it ever appropriate to "out-think" the client and deliver a system that they didn't ask for and don't know they'll be needing? How do you all approach this question? -----Original Message----- From: accessd-bounces at databaseadvisors.com [mailto:accessd-bounces at databaseadvisors.com] On Behalf Of Charlotte Foust Sent: Wednesday, September 10, 2014 11:44 AM To: Access Developers discussion and problem solving Subject: Re: [AccessD] Table Structure Ideas You need a table for samples with a field that allows you to determine the order. Then you might use a table for Fibers Morphology and another for Polarized Light Microscopy, each with the fields needed for the information that applies. Each of those tables would be child tables to the main Sample table, which would contain any other information collected for a sample. The Sample PK would be a FK in each of the child tables. That would allow you to present the information in a variety of layouts without straining yourself. This assumes a one-to-one relationship between the tables but could allow for the addition of multiple results if you added a unique key to each child table as PK and allowed duplicates of the parent key. If they wanted to add different kinds of results, you would just add another table. Charlotte On Tue, Sep 9, 2014 at 11:11 PM, ACTEBS <actebs at actebs.com.au> wrote: > Hi Everyone, > > > > Got a complex form that we're turning into a DB and having difficulty > coming up with a good table structure to suit the form and the way the > data is collected and input. > > > > They want the Access DB to mirror the form. You can see the form here: > > > > http://download.actebs.com.au/FormDoc.jpg > > > > As you can see by the example image above the sample are marked as A, > B, C, and D, but sometimes they display it vertically and other > horizontally, which is confusing the hell out of me. Any idea on how > best to design the table structure so the data is easy to work with down the track? > > > > Any ideas most welcome. > > > > Thanks in advance. > > > > Vlado > > > > > > -- > AccessD mailing list > AccessD at databaseadvisors.com > http://databaseadvisors.com/mailman/listinfo/accessd > Website: http://www.databaseadvisors.com > -- AccessD mailing list AccessD at databaseadvisors.com http://databaseadvisors.com/mailman/listinfo/accessd Website: http://www.databaseadvisors.com -- AccessD mailing list AccessD at databaseadvisors.com http://databaseadvisors.com/mailman/listinfo/accessd Website: http://www.databaseadvisors.com